Background
Visualization of skeletal elements is central to three-dimensional (3-D) kinematic analysis of joint motion. Indirect methods based on tracking surface landmarks (using reflective markers attached to the skin surface) within a calibrated volume (stereophotogrammetry) can contain artifacts (errors of transformation) due to integumentary displacements relative to actual skeletal motion [
1‐
5]. Direct visualization of bony elements during joint motion are typically accomplished
via fluoroscopy or cineradiography. Unfortunately, both of these techniques require the use of ionizing radiation, and outcomes from these techniques are restricted mostly to two-dimensional (2-D) analyses as the majority of these systems use single-plane imaging [
6,
7]. Emergence of the roentgen stereophotogrammetry (RSA) technique has enabled
in-vivo measurement of complex 3-D skeletal kinematics from a series of radiographs acquired with biplanar, orthogonal fluoroscopy [
1,
8,
9]. Although this technique is accurate, it commonly requires surgical implantation of markers in bones [
2,
9‐
12], although model-based RSA techniques have recently begun to appear in the literature [
13‐
15].
Recording a series of joint-motion images using x-ray fluoroscopy and then manually superimposing 3-D models of the same skeletal elements to match corresponding outlines in the x-ray images has been used to quantify
in-vivo joint motion. [
7,
16‐
18]. More recently, Gatesy
et al. reported using the scientific rotoscoping (SR) motion analysis technique, which involves biplanar fluoroscopy to image skeletal movements, creation of 3-D models of joint skeleton from high-resolution computed tomography (CT) scans, followed by model-to-image matching and registration (rotoscoping) performed over several frames of images yielding skeletal motion animation and 3-D kinematic data [
6,
19‐
21]. SR was developed from the X-Ray Reconstruction of Moving Morphology (XROMM) motion quantification technique, which tracks implanted markers digitized in biplanar fluoroscopic images captured within a calibrated imaging volume, instead of utilizing the model superimposition technique [
6]. Though accurate, both SR and XROMM techniques require corrections of geometrical distortions in images prior to the animation [
11,
12,
22‐
24]. While x-ray-based motion analysis techniques like SR, XROMM, and RSA are clearly novel and advanced, their translation to clinical research (i.e., human subjects research) has been limited due to health-related concerns associated with the radiation exposure [
25‐
29].
From a clinical research perspective, understanding
in vivo skeletal motion is of interest to both scientists and clinicians [
30‐
32]. More specifically, x-ray-based diagnostic imaging techniques measuring human inter-vertebral displacements have focused mostly on imaging the spine at static end-of-range positions [
32‐
40]. However, qualitative and quantitative assessments of spinal motion have been enhanced by quantitative radiographic techniques that track displacements of pre-assigned coordinate points of specific anatomic locations on orthographic spinal images and by real-time joint-motion evaluation with XROMM-like techniques (using CT/magnetic-resonance-imaging-based 3-D models) and RSA (with per-operative implanted vertebral markers) in human subjects [
22,
28,
38,
41‐
47]. Regrettably, these approaches still require exposure to ionizing radiation and, at times, require marker implantation on the bones.
Magnetic resonance imaging (MRI), when used for quantifying inter-vertebral motion, has mostly been restricted to the analysis of end-of-range sagittal-plane displacements [
48‐
50]. However, dynamic cine-phase contrast (cine-PC) or fast-phase contrast (fast-PC) imaging with ultra-fast gradient echo sequences have been employed for evaluating joint kinematics, especially in ankle, knee, or shoulder motion [
51‐
57]. The main approach for these techniques has been the use of pulse sequences that permit volume extraction from full 3-D motion datasets at selected time points along the range of motion (ROM). However, these techniques can be time-consuming. Additionally, the use of cine-PC sequences require a repeated, cyclic, velocity-controlled motion to be performed at the joint of interest during scanning to make the motion synchronized with velocity-encoded motion capture [
57,
58]. Also, these images have low resolution and may present motion artifacts [
56,
58,
59]. More recently, the combined use of segmented 3-D anatomical models (from high resolution, ~15 mins duration, static axial scans) registered to low resolution, volumetric images acquired at different joint positions using high speed (~40 sec) T
1 sequences has been reported [
57]. Although such techniques acquire multi-position data with much greater speed, the segmentation of these low-resolution images still require multi-slice images of the experimental quasi-dynamic joint positions. Accordingly, recent advancements in these methods have focused on the acquisition of faster and fewer slices of joint motion (without compromising image resolution) for model-to-image registration and without reducing the accuracy of the technique (time-accuracy tradeoff).
Currently, no modeling techniques exist for quantification of inter-vertebral joint displacements using single-plane or orthogonal magnetic-resonance (MR) image templates for 3-D model registration. Accordingly, our long-term goal is to develop a 3-D model-based technique that permits fast dynamic MR imaging of the human lumbar spine using an open-bore weight-bearing musculoskeletal MRI. Our study focuses on the lumbar spine as low back pain (LBP) is one of the most common reasons for seeking medical care world-wide and accounts for over 3.7 million physician visits per year in the United States alone [
60‐
64]. As such, LBP is arguably one of the most debilitating and costly health disorders, and the development of technologies to aid scientists and clinicians in better understanding the etiology of LBP—as well as in monitoring the effects of therapeutic interventions— is desperately needed.
As a first step towards our long-term goal, we present in this article our initial research and development findings for an MRI-only protocol involving imaging (using a standard T1 and a fast contrast-enhanced MRI sequences), a series of pre-determined displacements between solid-body models, 3-D models (segmenting T1 weighted axial scans), and a morphology-based rotoscoping strategy for animation and quantification of the displacements. The use of the contrast-enhanced sequence will allow us, firstly, to test the feasibility and reliability of its use as a fast imaging tool and secondly, to compare its outcome with that of the standard high-resolution T1 images. The feasibility and reliability of this MRI-based technique is discussed here, and we anticipate further developing this technique into a motion-assessment tool for the lumbar spine and other di-arthrodial joints.
Discussion
In this study, we describe a novel MRI-based approach that is conceptually similar to some fluoroscopy-based modeling protocols with the major difference of not requiring exposure to ionizing radiation, which has obvious implications for clinical research [
6,
18,
19]. While this is the first step in the development of an MRI-based protocol of this nature, our initial work indicates that this technique has promise as we have successfully developed a logical and rational approach to the quantification of motion and have also demonstrated relative and absolute reliability. Below we discuss our findings within the context of the extant literature as well as our future directions.
As stated above, the primary innovation of this work is that it represents an MRI-only, morphology-based modeling technique for tracking solid-body displacements, which is similar to fluoroscopy-based approaches, such as RSA, SR, and XROMM, and static x-ray-based techniques. The scope of application of these fluoroscopy-based techniques is limited due to the ionizing radiation exposure. For instance, obtaining serial measures involving significant radiation exposure over time in research studies requiring oversight by an institutional review board (or other analogous committees charged with approving, monitoring and reviewing biomedical research involving humans) could raise questions about the cost-to-benefit ratio, particularly in light of the Institute of Medicine’s recommendation on avoiding unnecessary medical radiation throughout life [
26,
28]. Accordingly, we believe that an MRI-only-based modeling technique for investigating joint kinematics has significant advantages, particularly for the advancement of clinical research.
Available MRI modeling techniques have usually applied multi-slice imaging of the objects of interest to capture the experimental displacements introduced into these objects in the scanning environment. Our study has uniquely attempted a morphology-based single-plane and an orthogonal imaging protocol to quantify experimentally induced displacements in our models. Additionally, we have used a fast-scanning protocol with dynamic contrast-enhanced pulse sequence and compared its outcome to a standard high-resolution T1 imaging. Both these methods have demonstrated high-levels of reliability in quantifying displacements in objects within the MR imaging volume. These findings provide basic proof-of-concept for the notion that a reliable non-ionizing-radiation-based motion quantification technique can potentially be used to characterize a quasi-static visualization of joint kinematics from a single and biplanar approach. The use of dynamic sequence and image processing can be further explored to attempt quantification of joint kinematics in synchronized motion. Additionally, while our single-plane technique does not objectively address detecting out-of-plane motion, inclusion of the orthogonal imaging in the biplanar approach helps manual positioning of the model to match the corresponding out-of-plane shifts of the image silhouettes.
While our initial development results are promising, our study has some limitations. First, we have used static two-dimensional imaging for quantification purposes; and we do not know whether comparable levels of reliability would have been observed if the dynamic pulse sequence were used to scan the solid-bodies in real-time during an un-synchronized motion with subsequent quantification of these images using the technique reported here. The approach we chose was based on technology currently available; to our knowledge, an MRI-compatible device that would permit real-time manipulation of motion is not commercially available, and the custom development of such a device would require significant resources. Second, the current approach required manual segmentation and post-processing, which is very time intensive. Accordingly, we do not know how the use of semi-automatic protocols or automatic iterative segmentation algorithms would have changed our results. Third, while we reported high levels of reliability for a novel biplanar imaging modality (i.e., quantification of motion in two planes, or coupled motion), the orthogonal images obtained for this analysis were not acquired simultaneously (i.e., an image slice was first acquired in one plane and then acquired for the corresponding orthogonal plane) due to the inherent limitation of MR imaging to do so. While this is not necessarily a limitation of the current work, it could pose a limitation for future work that seeks to acquire simultaneous multi-planar images of motion. Lastly, we only assessed reliability and did not assess accuracy. We are currently conducting experiments that will assess accuracy of our technique in a porcine spine model.
Competing interests
The authors declare that they have no competing interests.
Authors’ contributions
NM conceived the study idea and design, performed image acquisition, data analysis, drafting and revision of manuscript. SM supported in designing the study, development of technical strategy, draft of manuscript. JC provided inputs on data analysis and drafting of manuscript. SW helped with the study design and to draft and revise the manuscript. JT assisted with study design and in editing the manuscript. BC contributed with data analysis, major edits on original manuscript and the revision. All authors read and approved the revisions and the final version of the manuscript.